Gemini 2.5 Flash (gemini-2.5-flash) accepts text, images, video, and audio as inputs and generates text. But its API lifecycle matters for model selection: as of November 12, 2026, Google’s model-specific page says API access is limited to existing users until December 1, 2026. From that date, requests to gemini-2.5-flash will automatically route to Gemini 3.5 Flash-Lite and be billed at that model’s rates. Google also says the model is no longer selectable in AI Studio. So existing API users can plan around the stated routing date, but should not assume the original Flash model will continue serving those requests after it. (Gemini 2.5 Flash model page)
Google’s general deprecations page includes a broad statement that Gemini 2.5 models will continue through the API until further notice, but its stable-model table lists Gemini 2.5 Pro, not Gemini 2.5 Flash. For Flash’s specific API status and routing, the model-specific page gives the applicable model ID, existing-user restriction, and December 1 date. (Gemini API deprecations; Gemini 2.5 Flash model page)
Separately, Google Cloud documentation lists October 20, 2026 as the retirement date for Gemini 2.5 Flash support on Gemini Enterprise Agent Platform. That date has passed. The Cloud page distinguishes that platform’s retirement from possible Gemini API access under separate terms; it does not establish current post-retirement availability on the Cloud platform. (Gemini Enterprise Agent Platform model page)
Practical choice: If you already use Gemini 2.5 Flash through the Gemini API, treat December 1, 2026 as the stated routing change and check your account and billing before that date. If you are starting a project now, evaluate models currently available for new projects rather than planning on selecting Gemini 2.5 Flash. Google points new projects to Gemini 3.5 Flash-Lite or Gemini 3.8 Flash. If considering TokenForge Cloud, confirm directly whether it offers this model and how its routing and billing work; the available sources do not establish TokenForge support.
What Gemini 2.5 Flash does
Google lists the model ID as gemini-2.5-flash, with a 1,048,576-token input limit and a maximum output of 65,536 tokens. Text is both an input and output modality; images, audio, and video are input-only. That makes the model a candidate for workloads that need text responses based on mixed-media input, but the listed support does not mean it generates images, audio, or video. (Gemini 2.5 Flash model page)
Google also lists thinking, structured output, function calling, code execution, Google Search grounding, and URL context among supported capabilities. The model page marks image generation, audio generation, and Live API as unsupported. The Google Cloud feature matrix additionally lists computer use and the Interactions API as unsupported. Check the specific API surface and channel you plan to use; a feature listed for one channel should not automatically be assumed available on another. (Gemini 2.5 Flash model page; Google Cloud model page)
The headline input-token limit does not mean every media file can be sent through every route. Google Cloud’s specifications show route-dependent media and file limits, including exceptions for some non-migrated endpoints. Check the current limit for the modality, upload method, and endpoint you plan to use rather than relying on the context-window figure alone. (Google Cloud model page)
Availability depends on the channel
| Channel | What Google’s documentation says | What that means for selection |
|---|---|---|
| Gemini API | The model-specific page says API access is limited to existing users until December 1, 2026. After that, requests using gemini-2.5-flash automatically route to Gemini 3.5 Flash-Lite and are billed at its rates. |
Existing API users should plan for the stated routing and billing change. Requests may still return responses, but after the date they will not be served by the original model under that model string. |
| Google AI Studio | The model-specific page says Gemini 2.5 Flash is no longer available to select in AI Studio. | Do not assume it can be selected there for a new project. |
| Gemini Enterprise Agent Platform | The Google Cloud page lists retirement on October 20, 2026, a date that has passed, and distinguishes this platform’s retirement from possible Gemini API access under separate terms. | Confirm current availability for the specific Cloud product and deployment route; the captured page does not establish post-retirement status. |
| TokenForge Cloud | The available sources do not confirm whether TokenForge currently offers this model, which provider route it uses, or its pricing. | Verify availability, compatibility, and billing with TokenForge before designing around it. Do not assume its access or charges match Google’s direct API. |
Is it a fit for your workload?
Consider evaluating it if you have an eligible API path and need text responses based on text, image, audio, or video input, or need one of the listed tools. For example, a document-analysis service could assess text responses based on PDF or image input, while an audio workflow could assess transcription or summarization. These are possible use cases, not guarantees of accuracy, latency, or production suitability.
Evaluate another model or route if you are starting a new integration, depend on a capability Google marks unsupported, or cannot risk a change in model behavior or billing. From December 1, requests to the Gemini 2.5 Flash model string are scheduled to route to Gemini 3.5 Flash-Lite. Automatic routing does not establish behavioral equivalence between the two models.
A sensible evaluation is to run representative requests on the intended model or replacement and compare task success, structured-output validity, tool-call behavior, latency, and token usage. This is an evaluation plan, not a test result: no comparative test was supplied for this guide. Include inputs representative of your workload, such as large files or audio/video if those matter, and review outputs before changing production traffic.
Production API and cost checks
Before shipping, confirm the current API documentation for your route, the exact model string, authentication, supported request features, and response behavior. The captured Google Cloud page lists GenerateContent, Count Tokens, and Chat Completions as supported there, but it also lists that platform’s retirement date. Those details do not establish current support on another channel or on TokenForge Cloud. (Google Cloud model page)
No current general Gemini 2.5 Flash price is verified in the available official pricing excerpt. A third-party article reports $0.30 per 1 million input tokens and $2.50 per 1 million output tokens, but the supplied information does not establish that these rates are current or apply to your channel. Do not use them as a current quote or as TokenForge pricing. (Google Gemini API pricing; third-party price report)
For budgeting, obtain the current rate card for the chosen channel, currency, and billing mode. Then estimate token costs separately: input tokens ÷ 1,000,000 × input rate, plus output tokens ÷ 1,000,000 × output rate. Add other applicable charges or discounts only after confirming them for that route.
Bottom line: Gemini 2.5 Flash supports text output from text, image, audio, and video inputs, but its API path is time-limited: eligible existing users are told to expect automatic routing to Gemini 3.5 Flash-Lite from December 1, 2026, at that model’s rates. Check the relevant channel, test the intended model or replacement on representative requests, and confirm current costs before committing.
Sources
Gemini 2.5 Flash - Gemini API | Google AI Studio
Gemini deprecations | Gemini API | Google AI for Developers
Gemini 2.5 Flash | Gemini Enterprise Agent Platform